Research and Application on Improved Naive Bayesian Classifier Method

Among all the improving approaches of Naïve Bayesian classifier, integrated one-dependence estimators present their advantages both in accuracy and complexity. This paper proposes a new approach to weight the super-parent one dependence estimators. To verify the validity of the proposed method, the experiments are performed using 16 datasets collected by University of California Irvine (UCI) and made application on 5 Fault Diagnostic datasets of diesel engine. The comparison experimental results with other algorithms demonstrate the effectiveness of the proposed method.